Merge pull request #13 from bugparty/codex/研究代码与标准库速度差异

Add microbenchmarks and clean bounds check config

Author: bugparty

README.md

@@ -4,7 +4,7 @@ This repository provides a small bit vector implementation along with tests and
 
 ## Running the benchmarks without bounds checking
 
-Bounds checking is enabled by default. To benchmark without checks, configure and build with:
+Bounds checking is enabled by default. To benchmark without checks, configure and build with (this defines `BITVECTOR_NO_BOUND_CHECK`):
 
 ```bash
 cmake -S . -B build -DBITVECTOR_NO_BOUND_CHECK=ON -DCMAKE_BUILD_TYPE=Release

bitvector_benchmark.cpp

@@ -72,11 +72,88 @@ static void BM_Std_Access(benchmark::State& state) {
   }
 }
 
+
+static void BM_Bowen_SetBitTrue6(benchmark::State& state) {
+  size_t n = state.range(0);
+  for (auto _ : state) {
+    bitvector<> bv(n);
+    for (size_t pos=0; pos+5 < n; pos+=6) {
+      bv.set_bit_true_6(pos, 1);
+    }
+    benchmark::ClobberMemory();
+  }
+}
+
+static void BM_Std_SetBitTrue6(benchmark::State& state) {
+  size_t n = state.range(0);
+  for (auto _ : state) {
+    std::vector<bool> bv(n);
+    for (size_t pos=0; pos+5 < n; pos+=6) {
+      for (int i=0;i<6;++i) {
+        bv[pos+i] = true;
+      }
+    }
+    benchmark::ClobberMemory();
+  }
+}
+
+static void BM_Bowen_QSetBitTrue6V2(benchmark::State& state) {
+  size_t n = state.range(0);
+  for (auto _ : state) {
+    bitvector<> bv(n);
+    bv.qset_bit_true_6_v2(0, 1, n);
+    benchmark::ClobberMemory();
+  }
+}
+
+static void BM_Std_QSetBitTrue6(benchmark::State& state) {
+  size_t n = state.range(0);
+  for (auto _ : state) {
+    std::vector<bool> bv(n);
+    for (size_t i=0;i<n;++i) {
+      bv[i] = true;
+    }
+    benchmark::ClobberMemory();
+  }
+}
+
+static void BM_Bowen_IncrementUntilZero(benchmark::State& state) {
+  size_t n = state.range(0);
+  bitvector<> bv(n, true);
+  bv.set_bit(n-1, false);
+  for (auto _ : state) {
+    size_t pos = 0;
+    bv.incrementUntilZero(pos);
+    benchmark::DoNotOptimize(pos);
+  }
+}
+
+static void incrementUntilZeroStd(const std::vector<bool>& bv, size_t& pos) {
+  while (pos < bv.size() && bv[pos]) ++pos;
+}
+
+static void BM_Std_IncrementUntilZero(benchmark::State& state) {
+  size_t n = state.range(0);
+  std::vector<bool> bv(n, true);
+  bv[n-1] = false;
+  for (auto _ : state) {
+    size_t pos = 0;
+    incrementUntilZeroStd(bv, pos);
+    benchmark::DoNotOptimize(pos);
+  }
+}
+
 BENCHMARK(BM_Bowen_Set)->Arg(1<<20)->MinTime(5.0);
 BENCHMARK(BM_Std_Set)->Arg(1<<20)->MinTime(5.0);
 BENCHMARK(BM_Bowen_PushBack)->Arg(1<<20)->MinTime(5.0);
 BENCHMARK(BM_Std_PushBack)->Arg(1<<20)->MinTime(5.0);
 BENCHMARK(BM_Bowen_Access)->Arg(1<<20)->MinTime(5.0);
 BENCHMARK(BM_Std_Access)->Arg(1<<20)->MinTime(5.0);
+BENCHMARK(BM_Bowen_SetBitTrue6)->Arg(1<<20)->MinTime(5.0);
+BENCHMARK(BM_Std_SetBitTrue6)->Arg(1<<20)->MinTime(5.0);
+BENCHMARK(BM_Bowen_QSetBitTrue6V2)->Arg(1<<20)->MinTime(5.0);
+BENCHMARK(BM_Std_QSetBitTrue6)->Arg(1<<20)->MinTime(5.0);
+BENCHMARK(BM_Bowen_IncrementUntilZero)->Arg(1<<20)->MinTime(5.0);
+BENCHMARK(BM_Std_IncrementUntilZero)->Arg(1<<20)->MinTime(5.0);
 
 BENCHMARK_MAIN();

main.cpp

@@ -3,7 +3,10 @@
 #include <chrono>
 #include <iostream>
 #include <vector>
-constexpr size_t SIZE = 1000000000;  // 1 billion elements
+
+// Benchmark size: 1 billion elements
+constexpr size_t SIZE = 1000000000;
+
 // Benchmarks the performance of std::vector<bool> for setting, accessing, and traversing elements.
 void benchmarkStdVectorBool(){